AI Guides › Skills & Agents
Building A Skill Around The Task You Dread Most, Not The One That's Easiest
By Nigel Guy · 2 min read
The first skill most people build is the easy one — the task that was
already simple, that they could have scripted in ten minutes, that just
happened to be top of mind. It ships fast, it feels like progress, and it
saves almost no time at all, because the task it replaced was never the
bottleneck.
The rule: build your first skill around the task you actively put off —
if building it feels uncomfortable because the task is messy or
high-stakes, that discomfort is the signal you're pointed at the right
target, not a reason to pick something easier.
The mechanism: the Dread Inventory
- List the tasks you avoid, delay, or do reluctantly. Not the ones
you haven't gotten to — the ones you actively route around, or do
ten minutes before they're due because thinking about them earlier is
unpleasant.
- For each one, ask why it's dreaded. Usually it's one of a few
things: it's tedious but has to be done carefully, it requires context
you have to reconstruct each time, or it's emotionally uncomfortable
(a hard email, a difficult review) rather than technically hard.
- Sort by whether a skill actually helps with that reason. A skill is
very good at "tedious but requires careful, repeatable steps." It's
less good, on its own, at "emotionally uncomfortable" — though it can
still help by doing the mechanical parts, leaving you to do only the
part that actually needs a person.
- Pick the one where a skill removes the dread, not just the labour.
The test isn't "does this save time" — plenty of easy tasks save time
too. The test is "does this remove the reason I was avoiding it."
- Build it narrow, aimed at that one dreaded task, not a general
version of the category. A skill built to solve your actual avoided
task is more useful than a broader one built to look more impressive.
What to skip
Skip starting with the task that's easiest to demo — an easy demo proves
the tooling works, not that the skill was worth building. Skip building a
skill for a dreaded task before you've worked out why it's dreaded; a
skill aimed at the wrong reason (automating the mechanical steps of a task
that was actually dreaded because of a hard judgement call) will get built,
get used once, and then get quietly avoided again for a new reason.
Guardrails
- Not every dreaded task should become a skill. Some are dreaded because
they genuinely need a person's judgement every time, and a skill there
just adds a layer to route around.
- Building a skill for something you dread doesn't retroactively make the
underlying task well-defined. If you can't describe the steps clearly
enough to write the skill, that's worth knowing before you build one.
- Check back in after using it a few times — a skill that removes the
dread on paper but that you still avoid running is telling you something
the inventory missed.
All 751 AI guides · JulieMango plans from £17/mo